THE ST.

FRANCIS DAM
FAILURE
worst American civil engineering
disaster of the 20th Century
J. David Rogers, Ph.D., P.E., P.G., C.E.G., C.HG.
Karl F. Hasselmann Chair in Geological Engineering
Missouri University of Science & Technology
for the

Shlemon Specialty Conference

Dam Foundations Failures and Incidents
Denver, Colorado
May 16-17, 2013

 St. Francis Dam was a 205-ft high concrete gravity-arch dam

constructed by the City of Los Angeles between 1924-26
It failed catastrophically on March 12-13, 1928, killing at least
432 people, making it the worst American civil engineering
failure of the 20th Century

William Mulholland 1855-1935

Mulholland was Chief
Engineer of the Los Angeles
Water Co. from 1886-1902
Chief Engineer & General
Manager of the Los Angeles
Bureau of Waterworks &
Supply from 1902-1928
Principal visionary and
architect of the first Los
Angeles Aqueduct and the
LA Bureau of Power & Light

The Water Crisis of 1918-26

While the aqueduct was
under construction, the
Citys population grew from
284,000 to 425,000 people
Near record-low rainfall
beset the LA area,
beginning in the winter of
1918-19, lasting till 1924-25
Cultivation of the San
Fernando Valley increased
567% from 1914-1923
More water storage was
needed in the Los Angeles
vicinity for drought periods.
St. Francis was the largest
of 9 reservoirs built or
enlarged between 1920-26

Citys First Concrete Dam

Prior to 1923 all of the Citys dams had been constructed of

earth fill. Weid Canyon Dam above Hollywood was the
Citys first concrete structure, because insufficient clay or
water was available to construct a hydraulic fill
embankment dam. It was christened Mulholland Dam when
it was completed in mid-1925. The lake has always been
called Hollywood Reservoir.

Dam Site in San Francisquito Canyon

A construction camp had been established in San

Francisquito Canyon in 1911 during excavation of n
6.5 miles of tunnels in the Pelona Schist, between
the future locations of Powerhouses 1 and 2
Mulholland believed that the natural constriction of
the canyon was an ideal location for a dam

 Construction of the St. Francis Dam began in July 1924

with the construction of a 8-feet high cutoff wall at the
upstream heel. This view shows the placement of Outlet
#1 and the two steps lying beneath its point of
discharge.

 View looking upstream at the first forms being placed for

the upstream heel of the dam, against the 8 ft high
cofferdam wall. Note pillows of mass concrete and
absence of contraction joints.

Minimal Abutment
Excavation 1924-25

Views at left show the left abutment

excavation into the Pelona Schist
between 6 and 15 ft deep, while on the
right abutment the depth of excavation
averaged only about 4 ft deep.

 View looking upstream at the dam during construction. Note the

upper and lower concrete batch plants, the construction towers,
and the inclined troughs. Also not vertical cuts in the Vasquez
conglomerate for the access road.

Dam had a stepped downstream face

A unique aspect of both
the Mulholland and St.
Francis Dams was their
stepped downstream
faces
The width of each step
was unique to its
respective elevation,
varying between 5.5 feet at
elevation 1645 and 1.45
feet at elevation 1815
This figured prominently
in subsequent forensic
evaluations

Colorized image by Pony Horton

St. Francis Dam was

completed in May 1926
Volume was 130,000 cubic
yards of concrete
11 spillway panels were
fitted on the crest, each
18 high and 20 ft wide
Five 30-dia outlet pipes
had a maximum capacity
of 1184 cfs with full
reservoir
If all 5 outlets had been
opened at noon on Monday
March 12th , the reservoir
would have dropped only
1.67 feet by midnight

 The reservoir was brought up to within 3 inches of spillway

crest for the first time on Wed., March 7, 1928. All city
reservoirs were full by the following Sunday, March 11th.
Damkeeper Tony Harnischfeger called Mulholland on the
morning of the 12th to report spillage of dirty water from the
right abutment area. That morning about 2 cfs spillage was
coming over the spillway panels from wind-whipped waves,
shown here above around noon on March 12th.

Before and After

Comparative views taken of the dams upstream

face looking at the right abutment 12 hours before
the failure (at left) and the day after (right)
Note exposed keyway beneath right abutment dike

 The dam burst open at 11:57-1/2 PM on Monday evening March

12, 1928, sweeping everything in its path. The damkeeper, his
wife and 6-year old son lived mile downstream. They
became the floods first victims. One passerby noted lights
down in the canyon below the dam within an hour of the
failure. There were no lights on the dam.

 The dam went out in the middle of the nightand the

first 62 victims were employees and their families living
at Powerhouse No 2, 7300 feet downstream of the dam.
Lots and lots of dead children, such as the little boy
shown here, who was never identified.

Tidal wave of destruction

Just downstream of the dam the maximum depth of the

flood was about 125 feet, shown at left.
Almost a mile downstream the floodwaters spilled over a
natural saddle 120 feet above the channel
The average velocity in this reach was about 26 feet per
second, or 18 miles per hour.

 The initial flooding reached a maximum depth of 140 feet

filling all of San Francisquito Canyon. In the map shown here
the reservoir is indicated in dark blue and the flood limits in
light blue.

64 victims at Powerhouse 2

7300 feet downstream, the wave drowned 64 of the

67 city employees and their dependents, who lived
at Powerhouse No. 2, shown here.

PATH OF DESTRUCTION

The flood swept down San Francisquito Canyon

and inundated the SoCal Edison Saugus
substation, collapsing highway bridges at Castaic
Junction, then swamped the SoCal Edison
construction camp a few miles downstream at
Kemp, drowning 84 of the 140 workers camped
there.

INTO THE SEA

The flood reached the Pacific Ocean below Montalvo

around 5:30 AM. By this time the wave was only moving
about 5 miles per hour, but was two miles wide. Both the
railroad and highway bridges were washed out. A number
of bodies were recovered from the ocean, as far south as
the Mexican border.

 Charles Lees map of the flooded areas, extending 54 miles

from the St. Francis damsite, through Castaic, Camulos,
Filmore, Santa Paula, Saticoy and Montalvo. The
smoothed flow distance was 52 miles.

 Aerial oblique view of the dam site the morning

after the failure, taken by Spence Aerial Surveys.
Note the enormous landslide on the dams left
abutment, truncating San Francisquito Canyon
Road, at extreme right

 William Mulholland and his assistant Harvey Van

Norman view the dam site from several hundred
yards upstream the morning after the failure (left).
The massive void left by the left abutment
landslide is seen at right. Between 40 and 80
vertical feet of the dams left abutment was eroded
by the outpouring waters of the reservoir.

 Governor C.C. Young appointed a 6-man panel to investigate the

failure, which included two geology professors. They made a
single visit to the site on Tuesday, March 20th and depended
entirely on the information collected by others and transmitted
to them. Plane table survey of dam pieces by surveyor H. Wildy
shown at right.

 The geologists were impressed by the sharp contact

between the Pelona Schist and Vasquez formation
(misnamed the Sespe formation in 1928), along the old San
Francisquito fault (an inactive feature)
The panel suspected that hydraulic piping may have
occurred along the fault because the Vasquez beds were
subject to disintegration (slaking) when submersed in
water

 A colorized image of the St. Francis failure by Pony

Horton, showing the color contrast between the red beds
of the Vasquez [Sespe] conglomerate (at left) and the
grey colored Pelona Schist (at right)

Post-failure survey
of displaced blocks

State Division of Highways surveyor Horace Wildy

identified 11 of 20 concrete monoliths displaced by the
outbreak flood. The panel focused on Block 16 because it
was found further downstream than Blocks 11, 12, or 14.

The Missing Section

The portion of the dam between Blocks 2/3/4

and 5/6/7 (shown in yellow) was not identified in
the debris field until months later.
This was referred to as the missing section

FIXING BLAME
The Commission met
on March 19th and
issued their report 5
days later
They concluded that the
red conglomerate
underling the dams
right abutment was
unsuitable for a dam
foundation, and that the
failure began in that
area, along the old San
Francisquito fault

 One of the shortcomings of the plane table map of the displaced

blocks was that it did not include the relative elevations of the objects.
Blocks 12 and 14 from the base of the left abutment were located 26
feet higher than Block 16 and well off the right side of the channel.
Block 11 may have helped form a dam with Blocks 12 and 13 that
deflected subsequent flows off to the left, where Block 16 was found,
further downstream.

A chopped downstream toe

Charles Outland discovered inconsistencies with the Citys official

cross section when he examined this construction photo, which clearly
shows a chopped downstream toe, beginning at Elevation 1650. This
suggests that the base was ~152 ft wide instead of the 176 feet shown
on the design section given to the Governors Commission.

 Original (1923) design concept for the St. Francis Dam by the
LA BWWS, shown at left
The cross section given to the Governors Commission by
BWWS is presented on the right. It extends down to Elevation
1620. The red line approximates the actual limits of the dam.

Battered upstream face ?

The St. Francis Dam
appears to have been
constructed with a 1:27
and 1:10 batter of the
upstream face.
To date, no evidence has
been found to show that
the upstream batter was
increased to 3.5:10
below Elevation 1645, as
shown on the design
section given to the
Governors Commission,
dated November 1924.

Political promises

In June 1922 Mulholland promised the Citys Board of Public Service

Commissioners that one of his proposed reservoirs would store a
entire years supply of water for the City of Los Angeles south of the
San Andreas fault
Originally intended to be 180 feet high in May 1923, it was decided to
raise the dam 10 feet in July 1924, shortly after construction began.
Another 10 feet of height was added in July 1925
These changes raised the height of the dam by 11% without increasing
its base width, reducing the factor of safety against overturning

The Stevens Stage Record

A Stevens reservoir level recorder

was mounted on the crest of Block 1
(left view). It recorded a slight drop
of the reservoir beginning around 8
PM, then an increasingly sharp drop
beginning around 12 Midnight. The
timing mechanism may have been
slightly ahead of schedule.

Grunskys Ladder

San Francisco engineer Carl Grunsky discovered the crushed

remains of the wooden stage recorder ladder wedged in a
tension crack at the dams upstream heel. This testifies to the
heel having been in tension, which would cause cantilever
instability. Grunsky was an engineer of equal, or even greater,
renown as Mulholland .

The 1959 failure of the

Malpasset arch dam in
France pointed to the
vulnerability of concrete
arch dams to uplift,
especialy on steeplysloping abutments. In
most of the masonry dams
designed before the St.
Francis failure,
subdrainage was limited to
the maximum cross
section, and often ignored
altogether on the
abutments.

The main section of the St. Francis Dam was constructed

with 10 uplift relief wells set in two rows, as shown above.
This portion of the dam did not fail, only the sloping
abutments, which did not have uplift relief wells.

 Hydrostatic, or uplift forces act equally in all directions and serve

to reduce the effective weight of the dam, causing it to become
unstable.
If the dam tilted forward degree, this would explain the 3.67
inch drop of the reservoir, recorded 40 minutes before the failure.
When the dam cracked at its upstream toe, the resultant thrust
would have been shifted 240 feet downstream, promoting
overturning instability.

The development of full hydrostatic uplift was a controversial subject,

before and, especially, after the St. Francis Dam failure

Prior to 1928, engineering texts did not specify how to

account for, or mitigate uplift, in their design examples for
masonry gravity dams. These examples are from Smiths
Construction of Masonry Dams (1915) and Wegmanns The
Design and Construction of Dams (1917, 1922, 1927).

Design Methodology in early 1920s

The example is from
Folwells Water Supply
Engineering, 3rd Ed
(1926)

Prior to 1928, the example

designs presented in
textbooks summed the
gravity forces as a line of
thrust without reservoir
pressure and another line
of thrust with full reservoir
pressure.
Until 1945 most engineers
assumed that concrete was
sufficiently impervious to
resist complete saturation,
and that dams founded on
relatively impervious strata,
such as granite or gneiss,
would not be subject to
hydraulic uplift.

 Prior to 1928, most concrete gravity dams were analyzed

assuming the concrete to be perfectly dry. The dead
weight of the concrete was then compared to the
hydrostatic force of the water and see if the resultant
thrust, RT, fell within the middle third of the dams base.

 In 1945 Karl Terzaghi published an article which demonstrated that

water pressure could infiltrate mass concrete, saturating it.
A conventional analysis of cantilever stresses in St. Francis Dam
assuming full uplift reveals that the dam becomes unstable in
overturning when the reservoir rose to within 7 feet of its crest! Full
uplift may have developed beneath the sloping abutments, which
were not afforded uplift relief wells.

 The arch stresses on the St. Francis Dam became very

high when the reservoir was raised to within 11 feet of
spillway crest.
The dam was designed before the Trial Load Theory of
Arch Stress Distribution was developed, so it was not
designed to incorporate the contribution of arching to
its stability.

 Reservoir Stage Curve for St. Francis Reservoir

between March 1, 1926 and March 13, 1928

In 1926 the reservoir was filled 110 feet, up to elevation

1780 feet between June 1st and September 1st, then
drawn down about 20 feet through the fall and winter
months, when demand was lowest.

From January 5 to May 8,1927 the

reservoir was raised another 52 feet,
to elevation 1832, within three feet of
the spillway sills, and held there for 3
weeks, then drawn down to
elevations 1813 to 1819 ft, until
November 10th.

During the first year of operation several large tension

cracks formed transverse to the dams axis. These were
likely in response to the cement heat of hydration, which
would have been considerable for 130,000 yds3 of mass
concrete.

Four prominent contraction joints leaked noticeable volumes of

water in the main dam and required grouting

Several tension cracks formed in the concrete dike section

during the second year of operation, in 1927-28. These
began leaking noticeably in early March 1928.

During the high water stand of 1927 seepage increased

markedly through the downstream face of the dam.
Mulholland ordered the four prominent cracks to be caulked
with oakum, to prevent loss of cement grout injected into
these cracks.

The reservoir was raised to within three inches of the

spillway sill elevation of 1835 feet on March 2, 1928

This allowed full hydrostatic pressures

and uplift to develop in the
foundations.

Prominent
shrinkage
crack
observed
cutting
through
Block 5 after
the failure

The oakum caulking can be

discerned on the post-failure
images as dark lines across
the dams downstream face

70 Kv Power line went down at 11:57-1/2 PM

Southern California Edisons 70 Kv Borel Power Line

shorted out 2-1/2 minutes before Midnight. The tandem
poles supporting the power line were situated well above
the high water line, downstream of the dams right
abutment (shown, above right).
The Governors Commission missed seeing the
disposition of the downed lines because they visited the
site 7 days later, after the downed lines had been clipped
and restored by SCE crews.

The tandem 70 Kv Power line poles

were high above the dams left abutment
The tandem
power poles
were situated
well above
the maximum
reservoir
level, shown
by arrow at
left
It would
appear that
the power
line was
severed by
the left
abutment
landslide

The East Abutment Landslide

The east abutment landslide involved about

700,000 cubic yards of Pelona Schist, removed
in a short period of time. The landslide scar
extended 130 feet above the reservoir water
surface.

Map of the East Abutment Landslide

The reservoir water was only in direct contact with

about 25% of the landslide material that was
removed during the outbreak flood
The dams left abutment thrust against the center
of the landslide area, shown in brown

 The landslide of March 12,1928 was only a portion of a

much larger paleolandslide developed deep within the
Pelona Schist.

 Stanford Geology Professor Bailey Willis recognized the

significance of en-echelon tensile scarps that cut across
the Bee High Line Road, 200 feet above the reservoirs high
water line (see photo at right).
He drew the sketch at left showing relation of the 1928
landslide to a much deeper-seated paleolandslide complex,
developed in the Pelona Schist

 The prominent topographic benches developed on the Sierra Pelona

are relicts of enormous landslide grabens. At various intervals these
massive slides must have blocked San Francisquito Creek, creating
temporary landslide dams, which promoted the development of the
tree-filled glen on fluvial and lacustrine sediments that underlie the
old reservoir floor.

Landslides mapped along canyon

The areas outlined in red are paleolandslides developed within the

Pelona Schist during the past 100 ka.
The blue area is the outline of the St. Francis Reservoir when it
failed. Note how the toes of many of the paleolandslides were
inundated by the 1926 reservoir

Ancient Landslide Dam

A paleolandslide in vicinity of the St. Francis Dam site appears to

have dammed the creek during late Pleistocene time (the last 100,000
years). This dam constricted the canyon and created a much larger
reservoir than St. Francis, and the areal limits of the pool are shown
here. It also created favorable topography for a man-made dam

Citys block map

City engineers Ralph R.
Proctor, H.C. Gardett, and A.R.
Arledge made careful surveys
of the flood aftermath
Proctor had been BWWS
resident engineer on St
Francis, while Gardett and
Arledge were BPL engineers
supervising reconstruction of
Powerhouse 2
They identified the source
locations of 17 of the dams
displaced blocks
Two candidates for Block 35
were identified, one of which
was located the furthest
distance downstream

Block 35 found furthest downstream

Proctor, Gardett and Arledge identified two candidates for Block 35.
The one located furthest downstream is shown at right. Block 35
came from the base of the dams missing section, at the bottom of
the dams left abutment. Blocks 27, 28 and 35 were identified by
adhesions of schist on their base relative to original horizontality of
the concrete cold pour joints (pillows), which are easily discerned.

Like a Giant Jigsaw Puzzle

Many of the dams concrete blocks were located

downstream on the basis of their step widths and the
foundation material adhering to them (channel alluvium,
schist or conglomerate).

Rock Mechanics Evaluations

In the mid-1980s I began visiting the dam site to map the

geology; focusing much of my efforts on collecting
attitudes of foliation and jointing in the Pelona Schist.
These data were plotted up on stereographic projections
and input into computerized databases for subsequent
manipulation and evaluation.

Keyblock Theory

The Keyblock computer program sorts out the

combinations of discontinuity intersections which could
form complex blocks of varying form. In this area, three
interesting types of blocks were identified at the base of
the dams left abutment

Uplift of Wedge B at base of left abutment

Full reservoir pressure was applied to Rock Wedges A, B

and C identified at the base of the left abutment in the
Keyblock program.
These were found to lift upward, even under the weight of
the dam in that area (which diminishes rapidly
progressing up the left abutment)
Such lifting could have destabilized the dams left
abutment

Discontinuous Deformation Analyses

A joint-bordered element mesh was constructed of the

dams lower left abutment area, assuming rock
wedges A, B and C; in vicinity of Block 35. The dam is
represented by the rectangular block shown above.
The reservoir pool (water) is to the left of this block.

Results of the DDA Evaluations

The DDA simulations suggest that wedges A, B and C

would all lift significantly if subjected to pressure head
percolating beneath the dams sloping abutment. This
lifting could cause a catastrophic failure of the abutment
section, very similar to the mode of failure that
subsequently occurred at Malpasset Dam in 1959.

Malpasset Dam
The Malpasset arch dam
in France failed on its
initial filling in 1959
The dam was designed
by Dr. Andre Coyne, the
worlds foremost expert
in arch dam engineering
Dr. Pierre Londe spent
8 years unraveling the
failure mechanism
It was caused by
hydraulic uplift of a large
rock wedge lying beneath
the dams left abutment

Modeling a high pressure leak

The possible impact of high pressure orifice flow

emanating from the base of the left abutment was also
modeled, using Discontinuous Deformation Analysis.
This model predicted slope instability resulting from flow
emanating from the base of the missing section

Evaluation of
East Abutment Slide
Discontinuous Deformation
Analysis was used to evaluate
the East Abutment Landslide
This illustrates the destabilizing
effects of lateral loss of restraint
that may have triggered upward
migration of the slide mass
Large pore pressures may have
developed along the basal
detachment surface due to
entrapment in the mica schist,
greatly reducing inter-particle
friction
The reservoir extended about
halfway up the slope shown here

Dead Fish and the Sabotage Theory

After the failure thousands of dead fish were observed floating in deep
plunge pools excavated downstream of the failed dam. Some
suggested they must have been killed by dynamite
Autopsies of the dead fish revealed that they succumbed to silt
ingestion of their gills; suggesting a very turbid outflow
Based on later tests performed during the Second World War, it would
have taken more than 12,000 lbs of dynamite beneath 30 feet of water
on the dams upstream side to sabotage the structure

Back-analysis of the outbreak flood

The dam site could not be used for a back-analysis of

the outbreak flood hydrograph (quantity of flow versus
time), because the cross-sectional area was varying
with each passing minute of the failure; beginning
with a small area and concluding with the maximum
area, shown here.

Establishing constraining data points

We know from eye witness accounts that the reservoir

was essentially empty by 1:09 AM, as shown above
We know that the SCEs Lancaster power line went down
at 11:57-1/2 PM in vicinity of the dam site
Exactly five minutes later, we know that Powerhouse 2,
located 7300 feet downstream, went offline at 12:02-1/2
AM

Establishing constraining data points

We know the maximum depth of the outflow from the measurements

made at the time and used terrestrial photogrammetry to fill in the gaps
The depth of flood water was about 140 feet just downstream of the
dam and had diminished to about 110 feet at Powerhouse 2, 1.4 miles
downstream

Crucial data for a credible analysis

At Powerhouse 2, the surge chamber attendant E. H.

Thomas climbed down the tramway tracks during the
flood
He reached the high water line at 12:15 AM, and noted that
the level had already dropped 20 feet; shown by the
parallel blue arrows at left side of the photo at right

Reconstructed hydrograph at Powerhouse 2

A peak flow of 1.3 million cfs was calculated for

Powerhouse Two; 7,400 ft downstream of the dam
The peak flow at the damsite was likely close to 1.7
million cfs

Freighting of massive blocks

How were such massive blocks of concrete,

weighing as much as 10,000 tons, moved as much
as a half mile downstream?

Evidence of hydraulic sorting

Flood outwash particles are typically sorted in inverse

fashion; with coarse material on the bottom and
progressively finer material upward, because the stream
power of the flood subsides with decreasing flow.
Medium grained sand overlies coarse schist detritus (at
left)
The average particle size was 12 in photo at right

 The coincident excavation of 700,000 yards of schist with the

outpouring flood waters created an extremely turbid mixture.
As the percentage of entrained solids in the flood water
increases, the effective weight of the concrete blocks
diminishes to a fraction of their dry weight. This is how large
blocks were rolled so far downstream.

A casual
observation

En-echelon scarps across road

The Ray Silvey family was driving up San Francisquito Road past the
dam around 8:30 PM on March 12th
They had gone about 100 feet past the dams left abutment when they
were forced to stop by a 12-inch high scarp cutting across the road, in
the Pelona Schist!

Likely Failure Sequence

Several transverse cracks appeared in the dam during the

previous year, as sketched here
The entire left abutment (right side of this elevation view)
dropped 12 inches, at least 3-1/2 hrs before the failure
Some new leaks may have developed at the base of the
left abutment, out of view of the road above.

 High velocity orifice flow may have sprung from the base of the right
abutment, shortly before the failure.
A light was observed in the canyon below the dam by passersby in
the 45 minutes preceding the failure.
The damkeeper wifes body was found fully clothed, wedged
between two blocks near the base of the structure, mile upstream
from where she lived. This suggests the couple was up at the dam
looking at something.

Landslideinduced seich

Around 11:57-1/2 PM a massive landslide of the dams eastern

abutment initiated, severing the SCE 70 Kv Lancaster power lines.
The entirety of the dams left abutment was carried across the
downstream face of the main dam.
A landslide-driven displacement wave washed flotsam 4 ft above the
reservoir high water line, 3/4 mile to the north

 As the slide carried the dams left abutment

section across the canyon, the heavy blocks
sheared off 10 to 20 feet of the dams stepped
concrete face.

 The sheer size of Blocks 5 and 6 can be appreciated in this

photograph, with a person for scale (arrow)
Note the 5 to 10 feet cover of schist detritus preserved in
the steps of the block, 35 feet above the creek level.

 Another photo, showing the detachment of Block 6 from

Block 5. Note the four men for relative scale.

 Profile view taken

from left abutment
looking at the
massive cleavage of
the dams stepped
downstream face.
Note the angle of the
cleavage is tilted
downstream
Note the position of
Block 5, which slid all
the way to the right
abutment, then fell
about 35 feet (arrows)

Sheared face
of Block 1
Another photo,
showing the final
position of Block 7,
wedged between
Blocks 1 and 5.
Block 7 came from
the dams upper left
abutment, which slid
in behind Block 5.
Note the man at
bottom left for scale.

 The landslide debris dam is eroded by the outpouring

water over a period of probably less than five minutes
The out-rushing flow bent the cylindrical stilling well of the
Stevens Gage towards the left abutment
Block 5 originally turned upward, against the right
abutment

Rotation of Block 1 during outbreak flood

Five benchmarks were set in the crest of the dam

After the failure, the surviving stations were re-occupied and
it was determined that Block 1 rotated clockwise, with the
south edge shifting 8.4 inches towards the southwest, as
shown here
This rotation of Block 1 was likely caused by the outpouring
flood waters passing around the left side of the dam,
undercutting Block 2/3/4, before it toppled backward

 As the left side of the main dam was undercut, the dam
tilted slightly and rotated to that side, allowing water to
enter the shrinkage crack on the west side of Block 1
This triggered a chain-reaction failure of the right
abutment , but only after the reservoir had dropped
between 70 to 80 feet.

 Man standing on intact ground which was not inundated

by outpouring flood waters, at elevation 1765, 70 feet
below the reservoir water surface, along the construction
access road that traversed the dams right abutment.
The loose detritus in foreground suggest that the high
water mark may have been 5 to 10 ft lower.

 Looking towards remnant of the main dam from

the downstream right abutment, after the failure
No erosion can be observed to a depth of at least
70 to 80 feet below crest, suggesting that the
flooding did not initiate on this side of the dam

 Loose sidecast fill from construction access road

begins about 75 feet below crest (arrow)
Note schist detritus lying atop Block 5, 30 to 40
feet above the channel. This was left from the
landslide that Block 5 rode down, into the canyon.

 Towards the end of emptying of the reservoir, the left half

of the main dam topped backward at an angle of 54
degrees after being undercut by the outbreak flood waters.
The depth of this downcutting was about 35 vertical feet!
Patches of schist detritus were left upon Blocks 5 and 7
(shown in yellow)

 Photograph taken after the failure showing the disposition of

Blocks 1 thru 7. The left half of the main dam fell into an
enormous hole cut beneath its left side, between Elevation 1660
and 1615 ft, a depth of 45 feet!
Note position of Block 5, well below chatter line on downstream
face (arrows). Water pouring out of the right abutment breach
likely carried off the schist detritus beneath Block 5.

 The elevation of the

drained reservoir
pool at the time
Blocks 2/3/4 topped
backward (about 16
ft) was estimated by
studying the scour
marks left on the
tilted blocks and
comparing this
measurement with
the five-feet high
concrete lifts,
bracketed by the
blue arrows

The Final Configuration

The following morning this is how the dam site appeared.

Note the tandem SCE power poles well above the high
water line downstream of the right abutment (arrow)

 A sorrowful Mulholland told the Coroners Inquest

that he only envied those who were killed
He went on to say Dont blame anyone else, you just
fasten it on me. If there was an error in human
judgment, I was the human.

 The LA Co Coroners Inquest found that the

construction and operation of a great dam should never be
left to the sole judgment of one man, no matter how
eminent

So, who actually designed the dam?

Mulholland Dam was laid
out by BWWS office
engineer Edgar A. Bayley
(1877-1943)
No cores or tests of the
foundation rock were
undertaken
No formal calculations were
made
The design method used
followed the examples
presented in Smiths
Construction of Masonry
Dams (1915), Fowlers Water
The design for Mulholland Dam was simply
transferred to the St. Francis site, with a Supply Engineering (1926)
number of minor changes; nobody actually and Wegmanns The Design
claiming credit for having designed the and Construction of Dams
structure. This is the dams maximum section (1918, 1922, 1927)
in October 1925, after being raised the first 10
ft.

 Fixing blame: The Wednesday March 28, 1928 edition of

the Los Angeles TIMES ran headlines Foundation
blamed in Dam Disaster and Mulholland Takes Blame
for Mistakes

DESIGN DEFICIENCIES #1
The dam was unknowingly built against a
paleolandslide
Hydraulic uplift ignored in the design, leading to a
lower factor of safety than designers realized
Hydraulic uplift not relieved on sloping abutments (a
common problem until the 1960s)
Rather scant system of seepage interception
Cement heat of hydration effects ignored
Low strength laitance layer between successive
concrete lifts, creating low tensile strength horizons
Aggregate separation using trough placement

DESIGN DEFICIENCIES #2
The upstream heel of dam not battered 3.5:10 below
elevation 1645 ft
Dam heightened 20 feet without increasing base width
Downstream face chopped off at elevation 1650 ft, giving
a thinner cross section than it actually required, to
overcome the effects of uplift
Absence of grouted contraction joints
Plugging the dams expansion cracks with oakum on the
downstream face was the absolute WORST thing they
could have done to destabilize the dam
Gypsiferous Vasquez conglomerate subject to slaking
under submersion
No instruments placed within the dam structure to
monitor its actual performance

CONCLUSION No. 1
For every complex problem, there is a solution
that is simple, neat, and wrong
H. L. Menken

#1) A calculated Factor of Safety less than 1.0

does not, in of itself, mean that a structure
failed via the precise mechanism analyzed.
Various failure mechanisms compete with
one another, simultaneously. All manner of
failure mechanisms should be evaluated
without prejudice. This is difficult to do, for
we are all prejudiced by our lifes experiences

CONCLUSION No. 2
#2) We will not identify those
geologic features or structures for
which we are not specifically
looking for. We have to have in
mind what we are seeking,
realizing that we will seldom be
able to recognize those features
with which weve had little prior
experience

CONCLUSION No. 3
Engineering geology, by its nature, is a
very subjective science, built upon
each persons unique pedigree of
experience. The simple inclusion of a
geologist on a project, will not, in of
itself, insulate such projects from
disaster. Consider the fact that we now
know there are over 153 dams currently
existing in the United States which
were unknowingly built against ancient
landslides.

REQUIEM FOR MULHOLLAND

Like any person, Bill Mulholland
had weak points in his character.
His thirst for thriftiness was one of
these flaws, but that same trait
allowed Los Angeles to build its
municipal infrastructure AHEAD of
its burgeoning population, at rock
bottom prices
He had an enormous capacity for
innovation; getting difficult projects
completed on-time and on-budget.
Engineers of that era tended to
underestimate the complexities of
pore pressure response, especially,
on concrete dams
He had the depth of character to
accept responsibility for
shortcomings in the dams design
and construction which very few
people at the time fully
comprehended